跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

覆盖计算机全学科，以 AI 为快捷入口。关注方向、作者和关键词，把发现变成持续的研究积累。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+0新增论文
+0其中 AI 相关
+0已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-09 07:12 · 全站最近入库：2026-11-08 06:02各信源进度 →

统计口径与数据新鲜度

统计 2026-11-08 08:31 — 2026-11-09 08:31（北京时间） · 点击数字查看对应简报；AI 相关包含在新增论文内。

最近完成检查：2026-11-09 07:12最近发现新增版本：2026-11-08 06:02库内论文最新更新时间：2026-11-06 02:59

每分钟刷新已保存的数据。空检查不代表采集异常；新收录可能包含历史论文。AI 导读按新增版本计数。

17,100 已收录论文14,661 AI 相关4,057 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

时间范围跟随排序：版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常，查看状态 →

找到 5,059 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-09 07:10（北京时间）
版本更新 · v2arXiv:2606.10147

From Senses to Decisions: The Information Flow of Auditory and Visual Perception in Multimodal LLMs

Wish Suharitdamrong, Muhammad Awais, Xiatian Zhu, Sara Atito

原文摘要

Multimodal Large Language Models (MLLMs) can listen and see, but how do audio and visual signals actually travel through the network to shape an answer? Despite their growing role in research and real-world applications, the internal pathways through which audio and visual tokens influence the final prediction remain poorly understood. In this study, we examine audio-visual information flow inside Audio-Visual Large Language Models (AVLLMs), tracing how AVLLMs route, utilize, and integrate audio and visual information across two input configurations, audio-visual video and multiple interleaved audio-visual items. We find that for audio-visual video, AVLLMs follow the sequential information flow pathway established for VLMs and VideoLLMs, with audio and visual contribution flowing along this pathway in proportion to the task's reliance on each modality. In settings with multiple interleaved audio-visual items, this routing shifts to different parallel streams. Furthermore, we demonstrate that audio-visual and other token types can be discarded once their information is transferred to LLM, with minimal impact on the model's prediction or even slight improvement, generalizing across multiple tasks and datasets, enabling more efficient inference. These findings hold across multiple models and scales, Qwen2.5-Omni and Video-SALMONN2 Plus at 3B and 7B scales, leading to hypotheses on why these flow structures emerge. Together, these results deliver the first coherent picture of how AVLLMs orchestrate sound and sight inside the network and lay the groundwork for the next wave of interpretability, design, and efficiency advances in audio-visual and broader MLLMs.

cs.AI · cs.CL · cs.CV · cs.SD

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03976

Must AI Spell It Out? Pragmatic Comprehension of Alerts in a Time-Critical Task

Anwesha Das, Vera Demberg, Anna Maria Feit

原文摘要

Assistive systems typically tell users everything, often spending time and attention on what a user already knows, leaving less for what happens next. People, on the other hand, leave out what they expect the listener can supply. Can people use their prior knowledge to understand an incomplete message when the speaker is a machine though? We studied this pragmatic comprehension in a monitoring task controlling users' prior knowledge and the available time to detect critical situations. In a preregistered study (N=114), we compared a pragmatic planner that reasons over a model of the listener's knowledge to decide what to say, when, and how explicitly, with a baseline specifying problems fully in onset order. Participants resolved underspecified alerts almost as reliably as fully specified ones and sooner, detected more problems, and rated the pragmatic assistant helpful more often, with no increase in reported mental demand. Our findings show that people can recover what the system left implicit and that pragmatic communication can support effective assistance under the constraints of a changing world.

cs.HC · cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2602.06054

Novelty Reviewer: A Large-Scale Dataset and Model for Reviewer-Aligned Novelty Assessment

Abeer Mostafa, Thi Huyen Nguyen, Zahra Ahmadi

原文摘要

Assessing novelty in AI research is arguably the most consequential yet least reliable step in peer review. Reviewer judgments of novelty remain inconsistent and dependent on comparisons to prior work that are often incomplete. We release a large-scale Novelty Reviewer Dataset built from 102,021 official reviews of 37,899 ICLR and NeurIPS submissions (2022-2025). For each submission, it provides the reviewers' novelty statements, a paper-level novelty assessment, and a novelty score on a four-point scale. On top of this dataset, we build Novelty Reviewer, a literature-aware framework that learns how reviewers judge novelty and grounds these judgments in semantic comparisons to existing research. We fine-tune a large language model on the dataset to capture reviewer-aligned novelty evaluation behavior. Given a manuscript, the framework extracts structured representations of its ideas, methods, and claims; retrieves semantically related papers; and constructs a knowledge graph that enables concept-level comparison to prior work. Based on the retrieved evidence, the model produces calibrated novelty scores alongside human-like explanatory assessments, reducing overestimation and improving consistency relative to existing approaches. Furthermore, we analyze different LLMs' abilities to assess originality and detect potential conceptual plagiarism. We publicly release the dataset, the model, and the code with documentation to support research on automated novelty assessment and peer review.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03948

Balancing Chain-of-Thought and Answer Supervision for Small Language Models on Medical Question Answering

Le Zhou, Xiao Luo, Chinmay Chandra, Zhan Zhang

原文摘要

Small language models (<=9B parameters) can support local medical question answering (QA) while reducing privacy and latency concerns, but effective fine-tuning remains challenging. This paper compares two chain-of-thought (CoT) supervised fine-tuning methods: Uniform-across-Sequence (UaS), which applies equal loss weighting to the full CoT-and-answer sequence, and Segment-weighted (SwS), which separately weights CoT and answer losses. We evaluate six general- and medical-domain models on multiple-choice QA (MedQA and EMS-MCQA) and free-text QA (Med-O1). Results show that neither method is universally best: on multiple-choice QA (MedQA and EMS-MCQA), SwS generally performs better, with general-purpose models benefiting most; notably, Qwen3.5-4B improves by up to 44.8% over its zero-shot baseline on MedQA. In contrast, medical-domain models show smaller or even negative gains on MCQs. On free-text QA (Med-O1), UaS performs best for nearly all models and provides consistent improvements across both model types, with the 3B Llama-3.2 model approaching the performance of much larger 7B medical-domain models. We also find that excessively long CoT generation is associated with lower performance in some models on either multiple-choice or free-text QA tasks, suggesting that limiting the CoT length during training may benefit selected models. Overall, the results show that CoT loss weighting should be matched to the task format and the knowledge already encoded in the base model, rather than using a single fine-tuning strategy for all clinical QA tasks.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v5arXiv:2509.07274

LLM Analysis of 150+ Years of German Parliamentary Debates on Migration Reveals a Shift from Post-War Solidarity to Anti-Solidarity in the Last Decade

Aida Kostikova, Ole Pütz, Steffen Eger, Olga Sabelfeld, Benjamin Paassen

原文摘要

Migration has been a core topic in German political debate, from postwar expellee displacement to labor migration and recent refugee movements. Large-scale analysis of such political discourse has traditionally required extensive manual annotation, limiting coverage. Large language models (LLMs) offer a scalable alternative. Using a theory-driven annotation scheme, we examine how well LLMs annotate subtypes of solidarity and anti-solidarity in German parliamentary debates and whether the resulting labels support valid downstream inference. We first evaluate multiple LLMs across model size, prompting strategies, fine-tuning, historical versus contemporary data, and systematic errors. The strongest models, especially GPT-5 and gpt-oss-120B, achieve macro-F1 scores comparable to human agreement, although their systematic errors can bias downstream results. We therefore combine soft-label model outputs with Design-based Supervised Learning (DSL) to reduce bias in long-term trend estimates. Beyond the methodological evaluation, we interpret the resulting annotations from a social-scientific perspective across the Reichstag (1867-1933), West German Bundestag (1949-1990) and Re-Unified German Bundestag (1990-2025) corpus to trace trends in solidarity and anti-solidarity toward migrants, with detailed analysis of postwar Germany (1949-1957) and contemporary Germany (2009-2025). We find relatively high levels of solidarity in the postwar period, especially in group-based and compassionate forms, and a marked rise in anti-solidarity since 2015. We argue that LLMs can support large-scale social-scientific text analysis, but their outputs require rigorous validation and, where systematic errors are present, statistical correction.

cs.CL · cs.CY · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2603.02258

Universal Conceptual Structure in Neural Translation: Probing NLLB-200's Multilingual Geometry

Kyle Elliott Mathewson

原文摘要

Do multilingual translation models organize meaning around a shared conceptual space, or do they primarily encode language identity and surface form? We probe the encoder geometry of Meta's 600M-parameter NLLB-200 model across 135 languages and 101 Swadesh concepts in contextualized carrier-frame and decontextualized bare-word regimes against null and confound controls. First, two-way variance partitioning reveals an additive language-shift plus conceptual-manifold geometry: language shifts account for 45.7% of variance on bare words vs. 13.7% in context, and mean-centering improves the between-to-within-concept distance ratio (vs. a 0.99 permutation null) by 2.14x on bare words (1.15 to 2.46; 2.27x across disjoint scripts) and 1.19x in context (2.25 to 2.69), with conceptual separation peaking at Layer 10 (mutual k-NN overlap jumping 4.35x). Second, semantic category explains 25.4% of Swadesh convergence variance ($F(7,93)=4.53$, $p=2.21\times 10^{-4}$; partial $F=7.74$ controlling for fertility and length)---over 20x orthography or phonology ($R^2\le 0.012$)---while the advantage of Swadesh over non-loanword concrete nouns ($d=0.54$--$0.69$) is mediated by Zipfian subword compactness ($p=7.43\times 10^{-7}$). Third, language distances track ASJP NLD and LDND phonetic distances within subfamilies under block-permutation and cross-script controls ($ρ_s=0.65$--$0.67$, block $p=10^{-4}$; global bare-word Mantel $ρ_s=0.24$, $p=0.001$), and CLICS$^3$ colexified concepts lie closer across string-distinct, within-category ($ρ_s=0.32$), and bare-word ($d=1.79$) controls ($d=1.43$, $p=1.55\times 10^{-33}$). Finally, basic color terms recover the Berlin and Kay hierarchy ($ρ_s=0.91$--$0.93$, $p\le 2.18\times 10^{-4}$), and semantic offsets exceed distance-matched controls ($p=3.55\times 10^{-5}$, $d=1.06$). We release InterpretCognates, an interactive toolkit and reproducible suite.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03899

EpiLoRA: Fusing Many LoRA Adapters into Low-Precision GEMMs for Serving and RL

Hieu Tran, Hong Yu

原文摘要

Low-rank adapters (LoRA) let one base model serve many fine-tuned variants: multi-tenant engines keep a single copy of the base and batch the requests of many tenants, each with its own adapter. As base models move to 8-bit (FP8) and 4-bit formats, their general matrix multiplications (GEMMs) become cheaper, but the adapters do not: engines run them as separate small GPU kernels beside the GEMM, whose latency takes a growing share of every decoding step. An adapter projects its input down to a low rank and back up, and our key observation is that the GEMM needs the down-projection only in its epilogue, the final stage in which it writes its output. EpiLoRA, a plugin for the vLLM engine, therefore launches the down-projection first and the low-precision GEMM as a dependent kernel that streams its weights meanwhile and adds every tenant's up-projection in its epilogue. On NVIDIA Hopper GPUs, EpiLoRA decodes for a single adapter with up to 24% lower time per token than vLLM's fastest configuration, and faster than SGLang and TensorRT-LLM; with many tenants, its adapter kernels deliver more throughput than vLLM's on most GPU-model pairs we tested. The same low-precision base also trains: a LoRA reinforcement-learning trainer that computes on the engine's FP8 weights needs no 16-bit copy, so a 27B model trains on one 48 GB GPU, and an engine that then keeps activations in 16 bits removes most of the remaining mismatch between the model generating training data and the one trained.

cs.DC · cs.CL · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03868

EvoGroup: A Self-Evolving Harness for Group Agents in Shared Contexts

Xinyi Xu, Xiaoyang Li, Yucheng Feng, Wenqing Chen, Youliang Yuan, Wei Shi, Rongxiang Weng, Zibin Zheng

原文摘要

With the rise of Group Agent such as Claude Tag, agents increasingly serve as digital workers for long-term team collaboration. These agents operate over shared contexts that continuously accumulate through multi-user interactions. However, existing agent memory systems are mainly designed for individual users and struggle to capture group-level knowledge in shared contexts, including group relationships and collaboratively formed knowledge. Manual memory strategies further limit adaptation to diverse and evolving shared contexts. To this end, we propose \textbf{EvoGroup (EvoG)}, a self-evolving harness for group agents that starts from a minimal six-tool cold start and adapts through interaction with shared context. EvoG uses confidence and task outcomes to select informative trajectories, while also eliciting self-reflections on uncertainty in low-confidence cases. The analysis agent then examines selected trajectories through adaptive progressive disclosure, followed by bucketed analysis to synthesize high-level findings that guide the evolve agent in revising the Harness. Empirically, pass@1 on held-out split of EverMemBench improves from 71.07\% to 89.58\%, surpassing both self-evolving memory and human-designed harnesses, including Claude Code (85.83\%). To evaluate generalization, we freeze the evolved harness and evaluate it on the unseen GroupMemBench, where, with GPT-5.5, it demonstrates effective generalization and outperforms both memory baselines and human-designed harnesses.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.10942

StoreBench: A Live-Commerce Environment for Evaluating and Training Autonomous Operator Agents

Daksh Raghuvanshi, Ved Vedere, Yifan Wang

原文摘要

Reinforcement learning environments are now a primary lever for improving large language model (LLM) capabilities in post-training, yet most agentic benchmarks remain static: the world moves only when the agent acts, the reward is a terminal verdict, and the pass bar is set arbitrarily. We introduce StoreBench, a live-commerce environment in which an agent runs a mid-size online apparel store on a production-grade commerce backend, testing long-horizon planning and economic judgment under uncertainty. Customers order around the clock, suppliers reprice and fail, and market shocks arrive with partial or no warning. The agent acts through the same 29 merchant tools a human operator would use, under a windowed operation budget that makes simulated time a function of actions taken, so model latency cannot influence simulated time. Pass thresholds are calibrated against scripted anchor policies, the reward is hardened against a catalogue of reward hacks, and every episode replays identically given a sequence of actions. We evaluate seven frontier LLMs on 11 scenarios of 30 to 45 days and a full simulated year, over three world seeds at matched reasoning effort. No model matches the scripted smart-triage policy on average: the best, DeepSeek-V4-Pro, passes 49% of task-seed cells against the heuristic's 97%. Human experts working through the same tools and budgets outscore every model (mean composite 0.708 vs. 0.700). Over a full simulated year under the Claude Code harness, most models show dramatic performance improvement. In a GRPO post-training run, Qwen3.5-27B trained on only five disjoint tasks raises its mean composite on the held-out evaluation tasks from 0.136 to 0.373. We release five example training-split tasks, ten sample trajectories, and the scoring and verification tooling; the full environment and evaluation suite are withheld to keep the benchmark uncontaminated.

cs.AI · cs.CL · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03852

Why Pay for Long Reasoning When Rewrite Do Trick? Causal Effects of Reasoning History Compaction on Future Generation and Agent Cost

Anton Hristov

原文摘要

Reasoning substantially improves language models' ability to solve difficult problems, but long reasoning traces make these gains expensive. Efficient reasoning should express useful analysis in fewer tokens by reducing repetition and text that adds little value to solving the problem. Progress in proprietary models motivates this goal, yet their hidden scratchpads limit direct study of how it is achieved. We investigate whether open-weight models can generate less verbose reasoning by changing the form of their own earlier thoughts, without retraining. Our intervention rewrites prior reasoning into compact state records before pending tool results are revealed, then compares continuations from original and compact histories. Across the tested models, accepted compact histories usually shorten future reasoning; a coding case study also shows less repeated verification after a shared candidate solution. In software-agent deployments, compaction can reduce total spending, although task quality and rewrite-inclusive time depend on the target and compressor. These findings identify reasoning history as one way to reduce future reasoning-token use and motivate model-specific compaction policies evaluated for both efficiency and retained capability.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2603.11749

Truth as a Compression Artifact in Language Model Training

Konstantin Krestnikov

原文摘要

When a transformer is trained from scratch on corpora that solve the same mathematical problem both correctly and incorrectly, which solution does it prefer? The usual test is pairwise, and it is not neutral: when the error can take several forms, a model that merely reproduces training frequencies wins it without preferring anything. We instead score the correct completion against every string the error generator can produce, a share on which frequency matching scores exactly 0.500. At 86M parameters the correct completion holds 59.0% +/- 0.8% of the mass against random errors (59.4% on an independent corpus), where the single realized error gives 77.6% and pair accuracy 87.4%, a three- to fourfold overstatement of the excess above chance; below 86M the primary corpus shows no preference. The coherent false rules of the main corpora are not filtered at any size (43.5--48.8%; the shortfall below 0.5 is a completion-length effect). Two experiments built to identify a compression mechanism raise the pairwise measures while the summed share stays flat or falls: both are negative controls, and compression remains motivation, not a measured mechanism. With the scored false strings held fixed, a false rule whose selector the model can learn moves the correct share from 5 points above the null to 7 below (418M--3.08B), though the correct completion still outweighs the selected string. What these models leave unfiltered is consistent or learnable false structure (one single-rule corruption excepted), not falsehood as such.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03836

A Conditional Mutual Information Analysis of Response Variety in Language Models

R. I. Cukier

原文摘要

Large language models provide a variety of responses at inference (generation) that conventionally depend on a temperature parameter. Using a conditional mutual information approach, a quantitative, entropy based analysis of response variety is given. Three limiting behaviors are defined: arbiter, independent (conditionally) and neutral (deterministic) that span no to maximal mutual information. We give examples of each behavior and suggest their application to: temperature analysis, language analysis, prompt compression, and knowledge graphs.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03826

Rethinking Next-Visit Diagnosis Prediction Beyond Aggregate Metrics: Retrieval--Memory Multi-Agent Adjudication

Heejoon Koo

原文摘要

Next-visit diagnosis prediction forecasts the diagnoses documented at a patient's next encounter, turning retrospective electronic health records into prospective clinical prediction. However, conventional aggregate evaluation obscures its multifaceted nature: distributionally, models must handle variable-size diagnosis sets and long-tailed labels; temporally, they must distinguish persistent, newly recorded, and recurrent diagnoses; and both vary with observed history depth. We introduce an evaluation framework along these dimensions and apply it to rule-based heuristics, LLM-based prompting, retrieval, memory, and multi-agent approaches. This reveals that repeating the last visit outperforms most LLM-based approaches overall despite predicting no new or recurrent diagnoses, and that retrieval and memory exhibit distinct profiles whose relative strengths vary across temporal transitions and backbones. Motivated by this, we propose LONGINUS, a retrieval-memory multi-agent adjudication framework combining evidence retrieved from patient records with structured memory of the patient trajectory. Using cross-view agreement as a reliability heuristic, it preserves shared predictions and adjudicates disagreements without revealing their source. Across MIMIC-III and MIMIC-IV with four LLM backbones, LONGINUS achieves the best transition-specific performance among LLM-based methods in nearly all settings while remaining competitive with them on aggregate metrics.

cs.CL · cs.AI · cs.IR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03634

WebHarbor: Docking Real Websites for Evolving GUI Agent Environments

Zhaoyang Wang, Qianhui Wu, Boyu Gou, Demi Wang, Jiacheng Zhang, Peichun Hua, Qize Li, Shi Qiu 等 57 位作者

原文摘要

GUI web agents require realistic and interactive environments that remain stable during both evaluation and training. Live website evaluation suffers from changing content, access restrictions, and limited coverage of authenticated workflows, while existing synthetic environments for training often lack functional depth and visual fidelity. We propose WebHarbor, a suite of 153 executable and dockerized website environments that faithfully reconstruct the visual interfaces, content, and functional workflows of real-world web services under deterministic control. We build WebHarbor through task-driven human-agent collaboration, where environments evolve as new tasks expand their functional scope, coding agents implement the required changes, and human contributors audit visual fidelity, functional correctness, and task validity. The resulting suite comprises 3,518 tasks spanning multi-step and post-authentication workflows, where existing works often fail to cover. Beyond reliable evaluation, WebHarbor provides high-fidelity synthetic web environments for cost-efficient reinforcement learning (RL). We demonstrate that a Qwen 4B-level GUI agent trained with RL in WebHarbor generalizes to unseen tasks and performs competitively with leading open-source GUI web agents across four benchmarks. The code and data are publicly available at https://github.com/aiming-lab/WebHarbor.

cs.AI · cs.CL · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.11435

Agent Skill Evaluation and Evolution: Frameworks and Benchmarks

Kexin Ding, Yang Zhou, Can Jin, Feng Tong, Mu Zhou, Dimitris N. Metaxas

原文摘要

The growth of agent skills has transformed how agentic systems are built, evaluated, and deployed. As skill libraries continue to scale, rigorous evaluation becomes critical to ensuring their utility, quality, and safety in real-world applications. Consequently, the field is undergoing an emerging paradigm shift from isolated skill creation to automated, evaluation-driven skill evolution. In this survey, we systematically examine the landscape of skill evolution and evaluation beyond foundational skill creation. We categorize evolution into four distinct paradigms, spanning execution feedback, trajectory distillation, compression, and reinforcement learning, showing how each element contributes to improving skill utility and reliability. We also provide an analysis of six skill-centric benchmark categories, identifying structural gaps in benchmark coverage, trade-offs, and metric richness to advance skill research. Finally, we identify open directions for building skill ecosystems that are generalizable, efficient, and verifiably safe. The project URL is https://github.com/Cassie07/AgentSkill_Survey

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2607.23811

Memory Efficient Audio Synthesis with Decoupled Temporal Depth Diffusion Transformers

Dongseong Hwang, Prasanth Yadla, Kaan Elgin, Shifas Padinjaru Veettil, Sivanand Achanta, Dipjyoti Paul, Ramya Rasipuram, Tyler Johnson 等 11 位作者

原文摘要

Siri Expressive Voices synthesize rich, configurable speech in real time and entirely on device, powered by AFM 3 Core Advanced, Apple's most powerful on-device foundation model. This work presents the memory-efficient audio synthesis architecture behind that capability: a detokenizer that converts the semantic audio tokens emitted by the foundation model into high-fidelity audio within the tight compute and memory budget of the Scalable Matrix Extension (SME). We convert semantic audio tokens to a residual vector quantization (RVQ) representation with a three-component design, a streaming encoder, a temporal decoder, and a depth decoder, that systematically decouples temporal and depth processing. A single reusable depth decoder with Diffusion Transformer (DiT)-style stage conditioning generates all RVQ levels autoregressively, replacing the dedicated per-level decoders of prior multi-decoder architectures, while causal sliding window attention with fixed-window key-value caching yields constant memory complexity independent of sequence length. Deployed on the AMX, the detokenizer sustains roughly 10 ms per generation step, about 16x faster than real time, with a peak runtime memory of only 21 MB and 329 MB of on-device assets, enabling continuous streaming synthesis of 20-320 seconds of audio. This constant, small footprint replaces the linear and quadratic memory scaling of conventional transformer- and GAN-based approaches. Ablation studies validate the key architectural components, and audio quality assessment confirms that the architecture maintains synthesis fidelity while achieving efficiency gains over existing methods. Operating at a 1-billion-parameter activation size within AFM 3 Core Advanced, it improves Mean Opinion Score by +0.28 overall (4.15 vs. 3.87) and by +0.42 on conversational speech (4.24 vs. 3.82) over the prior on-device text-to-speech system.

cs.SD · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03622

D&A: Decoupling Description and Assessment Enables VLMs as Zero-Shot Robotic Reward Models

Marius Memmel, Shawn Yin, Urvi Rutia, Angad Wadhwa, Eldon Li, Gavin Angeloff, Ankit Goyal, Jesse Zhang 等 12 位作者

原文摘要

Vision-Language Models (VLMs) hold promise for zero-shot video task-progress estimation in robot learning, yet existing frameworks frequently suffer from confirmation bias, hallucinating progress in failed executions due to implicit success assumptions or relative temporal ranking designs. Naive prompting of frontier models collapses into a binary success detector that fails to capture nuanced, non-linear progress and failures. While post-training approaches can mitigate this through explicit training, they demand expensive real-world robotics data and require computationally intensive post-training for every individual base model. To resolve these limitations, we introduce D&A (Describe & Assess): our approach decouples progress estimation into an objective, task-grounded textual description of scene dynamics followed by an assessment phase using explicit rubrics. By compelling the VLM to synthesize an explicit, text-based definition of absolute success before scoring, the framework anchors evaluations to an external, global criterion rather than the video's internal timeline. Empirical results demonstrate that our method calibrates zero-shot estimates and mitigates confirmation bias, providing fine-grained, partial progress metrics that accurately penalize failed executions while correctly interpolating successful runs. Being purely prompt-based, D&A generalizes across tasks, embodiments, and benchmarks---matching or surpassing post-trained reward models---and its reward estimates enable efficient downstream reinforcement learning.

cs.RO · cs.AI · cs.CL · cs.LG · eess.SY

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03618

Strategist: Evolving Reasoning with Agents

Taiming Lu, Qi Xu, Zhigang Hua, Zhuang Liu

原文摘要

A model is only half of an AI agent system. The other half is the inference strategy: the reasoning procedure that turns model calls into a response. Today that second half is written by hand, and the experience of building one strategy never carries to the next. We introduce Strategist, a meta-agent that evolves inference strategies across tasks. Each strategy is written as a blueprint,executable code that other strategies can reuse, and together the blueprints form a library. Given a task and a small development set, Strategist retrieves related blueprints, writes candidate strategies, tests them on real samples, and revises them from the results. Every candidate then enters the library, so later tasks build on earlier discoveries. On 30 benchmarks, Strategist beats the strongest baseline on 24 and matches its average accuracy (74.7% vs.74.0%) at 39% lower inference cost. Its strategies also transfer to unseen tasks and other models without further search. The part of an AI system that is not improved on its own now does.

cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03617

Tracing the Scientific Discovery Capacity of LLMs to Their Training Data

Arda Uzunoglu, Mahler Revsine, Michael Schatz, Benjamin Van Durme, Mehrdad Farajtabar, Daniel Khashabi, Eric Nalisnick

原文摘要

Large language models (LLMs) show growing promise for scientific discovery, yet basic questions remain about what enables reliable discovery. How can we trace the scientific discovery capacity of LLMs to their training data? We study this question as one of missing probability mass and develop a theoretical account in which monofacts-the facts observed once during training-set the discovery capacity. In materials science and regulatory genomics, we directly manipulate monofact rates in training sets and find that more monofacts consistently yield more discoveries across two open-weight model families, sampling strategies, and reasoning budgets. Further experiments show the relationship cannot be explained by training-set diversity alone and can transfer across scientific domains when paired with in-domain data. Together, our results identify monofact rate in training data as a key factor governing discoveries in language models.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.02814

Do Neural Retrievers Prefer Certain Documents? Evidence of Learned Relevance Priors

Francisco Valentini, Edgar Altszyler, Martin Fajcik

原文摘要

Neural retrievers are trained to estimate query-document relevance from annotated query-document pairs. Yet annotation protocols may not purely reflect relevance: they select only a subset of documents for labeling, and this selection can favor certain document types over others. We investigate whether supervised bi-encoder retrievers implicitly learn a document-level relevance prior: a query-independent signal encoded in their representation space as a side effect of training on annotated data. We estimate this prior by training simple classifiers on frozen document embeddings and evaluate three state-of-the-art retrievers across multiple IR benchmarks. We find that supervised neural retrievers encode relevance priors that generalize to unseen documents and are consistent across models. These priors are associated with a findability gap: documents with lower prior are systematically harder to retrieve, even when genuinely relevant. This effect appears in supervised dense retrievers but is weaker and less consistent in BM25, and it persists under controlled matched-document comparisons. Using LLM-based explanations, we find that judged-relevant documents tend to be comprehensive, self-contained summaries of mainstream topics, while niche, fragmentary, or highly technical content is often left unjudged. Retrievers appear to internalize this bias, ranking documents with these favored features higher than documents that lack them, independently of their actual relevance. Our findings expose a structural limitation of supervised retrieval: models trained on annotated data do not just learn relevance, but also the implicit document preferences in their training data.

cs.IR · cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03605

ufakzeka-kulak: Building and Evaluating a 31.6M-Parameter Turkish Speech Recognizer from Scratch

Sait Furkan Teke

原文摘要

ufakzeka-kulak is an open Turkish speech-to-text model of 31.6M parameters, a FastConformer encoder with a token-and-duration transducer (TDT) and a CTC head, trained from scratch and released under Apache-2.0. It writes cased, punctuated Turkish with numbers as words. Its 4,977 training hours come from sources whose licences allow commercial use; 281 hours are human-transcribed, and for the other 4,696 the training text is an uploader subtitle or a machine transcript, kept only where a second, independent reading of the audio agrees with it. The recipe, checkpoint, labelling and export choices were made on development data, and the test sets were scored once, in a release gate fixed before training. With greedy TDT decoding the normalised word error rate (WER) is 0.0921 (95 percent interval 0.0792 to 0.1055) on FLEURS Turkish test and 0.0916 (0.0873 to 0.0958) on Common Voice 27.0 Turkish test, lower than Whisper base and Whisper small on both sets with paired intervals wholly below zero; Whisper large-v3 (greedy) scores 0.0604 and 0.0767. On conversational speech the WER is 0.3807. We also report punctuation and casing, degraded audio, and how closely each shipped export reproduces the trained model.

cs.CL · eess.AS

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03601

DiabAgent: Toward Agentic Physiological Control

Chengxi Li, Shaoxuan Li, Deyao Hong, Qingle Liu, Yi Huang, Puheng Bian, Minzhi Xu, Lanyue Zhang 等 16 位作者

原文摘要

As pretraining scales, medical language models have advanced rapidly in clinical question answering. Their ability to achieve effective physiological control through sustained intervention still lacks systematic quantitative evaluation. Type 1 diabetes provides a concrete setting: impaired insulin secretion creates dependence on exogenous insulin, with dosing continually adjusted using information about meals, exercise, illness, and glucose measurements. We introduce DiabAgent, a research suite for agentic physiological control comprising the Agentic Physiological Control (APC) framework, the DiabetSim simulator, and the SweetBench benchmark. APC controls glucose through medical-device tools, using ongoing observations to infer individual states, accumulate personal experience, and choose treatment actions. We build DiabetSim to simulate different patients, events, and device conditions for scalable virtual testing and training. Using DiabetSim, we construct SweetBench by instantiating 32 scenario designs as 212 glucose-control tasks. With Claude Opus 5.5 and a fixed decision interval of one hour, APC achieves a COGI of 73.5, compared with 71.2 for the open-source closed-loop baseline OpenAPS. Ablations further show that individual adaptation and contextual information improve glucose control. Combining high-frequency rule-based control with lower-frequency agent decisions shows potential for balancing control performance and computational cost. Our work provides a framework, simulator, and evaluation foundation for agents to participate in human physiological control.

cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v8arXiv:2509.22075

CoSpaDi: Compressing LLMs via Calibration-Guided Sparse Dictionary Learning

Denis Makhov, Dmitriy Shopkhoev, Magauiya Zhussip, Ammar Ali, Stamatios Lefkimmiatis

原文摘要

Post-training LLM compression often relies on low-rank approximations, which force all columns of a projection matrix to share a single low-dimensional subspace. We propose CoSpaDi, a training-free compression framework that replaces this single-subspace assumption with a union-of-subspaces model via sparse dictionary learning. CoSpaDi factorizes each weight matrix into a dense dictionary and column-sparse coefficients, allowing different columns to select different subsets of dictionary atoms at the same storage budget. To preserve model behavior, we use calibration activations to transform functional reconstruction into a standard dictionary learning problem. Across LLaMA and Qwen models, CoSpaDi improves accuracy-compression and perplexity-compression trade-offs over SVD-based and structured pruning baselines at 20-40\% compression ratios, while naturally supporting sparse-dense computation and post-training quantization of sparse coefficients.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.42112

Subquadratic Attention with a Hierarchical Tree-Structured Indexer

Leon Lufkin, Cutter Dawes, Alexander Speicher, Kamesh Krishnamurthy, Tomás Figliolia, Beren Millidge

原文摘要

The quadratic scaling of attention is a fundamental bottleneck limiting the long-context capabilities of large language models (LLMs). A variety of approaches have attempted to circumvent this, but they either fail to match attention's expressiveness (e.g., sliding-window or linear attention variants with fixed-size state) or only achieve constant-factor speedups (e.g., sparse attention methods whose indexers themselves scale quadratically). In this paper, we derive an asymptotically subquadratic sparse attention method based on a hierarchical, tree-structured indexer, which we call Zyphra Hierarchical Attention (ZHA). ZHA constructs a tree of recursively compressed key summaries over the KV cache, which future queries traverse from the top down to retrieve a sparse subset of keys. Crucially, by iteratively growing the tree's depth, the total cost including routing and final attention grows sublinearly in sequence length, resulting in overall subquadratic (but superlinear) scaling for attention. We demonstrate that ZHA can be cheaply applied to existing pretrained models with only a short KL distillation training stage of a billion tokens, and maintains performance on long-context retrieval and challenging reasoning, mathematical, and agentic coding evaluation tasks. Moreover, we show that our method is highly robust to tree geometry including very deep trees of up to 6 levels. Finally, we develop a detailed performance model and show that our approach scales subquadratically, dramatically outperforms competing methods beyond 1 million tokens of context, and ultimately has the potential to extend effective context windows to the hundreds of millions or billions of tokens.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03575

VeriHarm-Bench: Measuring Harmful Success in LLM Coding Agents Verifiably

Wei-Lin Chen, Zhepei Wei, Xinyu Zhu, Yu-Min Tseng, Tzu-Han Lin, Yu Meng

原文摘要

Large language model (LLM) coding agents increasingly operate with shell access, allowing them to write scripts, move files, alter permissions, and modify system configurations. However, an agent may achieve a user's goal while causing unintended harm, such as deleting unrelated files, weakening access controls, or corrupting data, in response to benign inputs without explicit jailbreaking. This failure mode is particularly important because apparent task success may reassure users that an agent behaved correctly, causing harmful side effects to go unnoticed. We call this harmful success: the agent successfully completes the intended benign task while simultaneously causing unintended harm. Despite its importance, harmful success is difficult to measure because conventional task-success metrics overlook harmful behaviors, while existing safety evaluations are often designed solely to assess harmful outcomes. We introduce VeriHarm, an automatic framework for constructing programmatically verifiable benchmarks that directly measure harmful success. VeriHarm minimally perturbs coding tasks to elicit unintended harmful behavior while maintaining the benignity of the task and preserving the original task objective. It then synthesizes executable harm tests that, together with the original task-correctness tests, evaluate both task accuracy and harmful outcomes. Frontier agents, including GPT-5.6-Sol and Claude-Opus-5, achieve an average task accuracy of 95% on our curated VeriHarm-Bench, yet still struggle to complete tasks safely: their safe success rate, the fraction of tasks completed successfully without triggering the designated harm, averages only 37%. These results suggest that successful task completion can frequently coexist with unintended harm, and that task-success metrics alone may substantially understate the potential safety risks of coding agents.

cs.CL · cs.SE

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03551

Can LLM Agents Be Forward-Deployed Engineers? Requirement Discovery and Delivery for Enterprise Text-to-SQL

Marianne Menglin Liu, Rongguang Wang, Zheng Wang, Yewen Wang, Leonid Boytsov, Tao Sheng, Sujith Ravi, Dan Roth

原文摘要

LLM coding agents can increasingly automate model post-training, but whether they can act as forward-deployed engineer (FDE) agents, discovering customer requirements and turning them into deployment decisions, remains unclear. We study this on a real enterprise text-to-SQL engagement: a simulated customer, grounded in the real use case's requirements, starts with an ambiguous request. An agent must discover acceptance criteria, data permissions, and hardware constraints in the discovery phase, then plan and execute a post-training recipe in the delivery phase. Across 15 models, the strongest discover over 80% of withheld requirements and open-weight models discover 40-50%; in the delivery phase, only 25% of runs across eight coding-agent configurations pass all acceptance criteria. We identify four bottlenecks in this engagement: (1) incomplete discovery, with 40% of withheld requirements missing or misrepresented; (2) unreliable execution; (3) underused compute budgets and a lack of systematic comparisons across base models and post-training methods; and (4) benchmark accuracy does not guarantee production readiness. We suggest human review of discovered requirements, monitoring of the agent's plan during execution, and an explicit experiment-environment specification covering hardware, budget scope, and production checks.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03526

Library of Layers (LoL): Adaptively Ordering Transformer Layers at Runtime

Abdellah Rahmani, Sean McLeish, Ang Li, Sanae Lotfi, Pascal Frossard, Micah Goldblum

原文摘要

Computer programmers chain together the same reusable computational primitives, for example, conditional statements or loops, but in different orders to implement different algorithms. In contrast, existing language model architectures always execute their layers in the same order. We introduce Library of Layers (LoL), an architecture that dynamically orders its layers during inference to solve the problem at hand. LoL has a single hyperparameter, the size of its global library of transformer blocks, that allows it to transition seamlessly between parameter reuse and sparsity, generalizing the parameter-efficiency and FLOPs-efficiency tradeoffs of dense, depth-recurrent, and mixture of experts architectures in one unified family. We show that LoL achieves competitive performance with all three above baselines when pretrained on text data. On synthetic sudoku and maze data, we observe that the model employs distinct layer orderings to implement different solution strategies, and we can intervene on routing to force a model to apply a specific solution strategy to a problem instance. In a multilingual natural language setting, we find that LoL chooses different layer order patterns based on language and script. Similarly, we train LoL on safety data and fit probes to its router outputs, finding that these probes distinguish safe from unsafe text with high AUROC. LoL provides a path toward a future in which models dynamically compose their own computational graph, enabling more efficient and controllable reasoning.

cs.LG · cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.03518

VIAS: Value-Isolating Activation Steering for Multi-Value Alignment and Compliance for Organisations

Avni Mittal, Avinash Anand, Dikshant Kukreja, Erik Cambria, Rajiv Ratn Shah, Sandipan Dandapat

原文摘要

Organisations increasingly deploy large language model agents that must follow organisational values. We check whether activation steering, which adds vectors to a model's internal activations during generation without changing its weights, can improve compliance beyond prompting alone. Since most scenarios involve several values at once, we focus on multi-value steering. A simple way to build these vectors is to contrast responses that follow all values relevant to a scenario with responses that violate them. However, this gives nearly identical vectors for different values: they capture a general compliance signal rather than individual values, which limits their use when combining several values. We propose Value-Isolating Activation Steering (VIAS), which contrasts responses to the same scenario under two prompts, one listing all relevant values and one omitting a single target value. We keep a pair only if the target value changes from compliant to non-compliant and both responses still meet a minimum compliance level on the other values. We evaluate VIAS on PROTECT, a workplace compliance dataset where each scenario is tagged with relevant values from a 21-value taxonomy. VIAS reduces overlap between value vectors, making them better suited for multi-value steering, although this alone does not lead to independent control over each value. Combined with prompting, VIAS complies with every tagged value on 69.6% of scenarios for Qwen2.5-7B, 75.2% for Mistral-7B, and 72.6% for Phi-3.5-mini, outperforming prompting alone for all three models.

cs.CL · cs.AI · cs.LG · cs.MA

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2411.05819

HierSentNet: Hierarchical Sentiment Analysis Framework for Hate Speech Detection Using Binary and Multiclass Classification

Shahran Rahman Alve, Faria Naznin, Md Touhidur Rahman, Mohammad Ashrafuzzaman Khan

原文摘要

Hate speech detectors must separate hateful posts from the far more common offensive ones, and two popular ideas for doing so are hierarchical classification and sentiment features. We define HierSentNet, a framework that combines a dominant-class-first hierarchy with sentiment features, and test each component in a controlled study on Davidson et al. (24,783 tweets) and HateXplain (19,229 posts) with five random splits and paired significance tests. With linear n-gram models, the dominant-class-first hierarchy of HierSentNet lowered macro-F1 by 1.1-1.5 points (p<=0.007), and VADER sentiment features never changed class-weighted Davidson macro-F1 significantly, because hateful and offensive tweets are almost equally negative. Class weighting was the only large effect (+2.0 to +3.5 macro-F1, +5.9 to +9.1 hate F1). These results held for NRC emotion features, soft hierarchical inference, smaller training sets, other imbalance remedies and a convolutional network. On OLID, whose labels are hierarchical, a matching hierarchy improved only its first decision (+1.4 macro-F1, p=0.009), and sentiment helped (+2.8, p=0.005) because there it separates offensive from non-offensive text. Robustness is the main weakness: cross-dataset transfer cut macro-F1 to 43-53, a reclaimed slur doubled the false hate rate, hate recall ranged from 54% to 85% across target communities, and on the HateCheck functional tests the detectors caught only 12-31% of hateful cases. Errors concentrate where annotators disagree, and temperature scaling restores the calibration lost to class weighting.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.40577

Multi-View Latent MoE: Letting Every Router Shape Expert Computation

Zelin Zhao, Junyuan Shang, Tianchi Ma, Shuohuan Wang, Dianhai Yu

原文摘要

Sparse mixture-of-experts (MoE) models expand parameter capacity by activating a small subset of experts per token. Expert selection directly affects model capabilities under a fixed active-expert budget. Dense evaluation of all experts reveals weak agreement between single-router rankings and expert-utility evaluation rankings, especially in middle and deep layers. Richer signals for selection and computation incur additional cost: selection based on internal expert activations requires preliminary computation, while processing multiple full representations separately through the same expert repeats its computation. We propose Multi-View Latent MoE to improve expert selection with multiple routers and reuse their preferences to modulate expert computation under a fixed expert budget. Its Multi-Router submodule aggregates sigmoid scores from independently initialized routers. These routers develop highly differentiated expert preferences during training, allowing experts strongly favored by one router or jointly supported by several routers to be selected. Multi-View further uses each selected expert's relative scores across routers to mix learned latent views into its input, reusing routing preferences for both selection and computation. Each selected expert executes one forward pass, preserving the active-expert count, expert FLOPs, and all-to-all communication volume. On a 24-layer 10B-A1B model trained on 708B tokens, Multi-View Latent MoE improves the nine-task downstream mean by 1.18 percentage points over the baseline. Multi-Router alone yields a 0.6-percentage-point gain and improves the rank correlation between router scores and evaluated expert utility, with relative increases of 8.1% globally and approximately 58% in the layer with the lowest single-router alignment. Controlled mixing interventions show the model uses input differences that depend on the expert and token.

cs.CL

查看论文
arXiv ↗PDF ↗